A quantitative structure activity relationship model for predicting minimum ignition energy of organic substance. (September 2020)

• Record Type: Journal Article
• Authors: Chen, Hsu-Fang; Chen, Chan-Cheng
• Is Part Of: Journal of loss prevention in the process industries. Volume 67(2020)
• Publication Date: 2020-09
• Subjects: Minimum ignition energy -- Quantitative structure-activity relationship -- Molecular descriptors; Chemical industries -- Safety measures -- Periodicals; 660.2804
• DOI: 10.1016/j.jlp.2020.104227
• Languages: English
• ISSNs: 0950-4230

Abstract

Abstract: Due to the high experimental cost and the danger in conducting tests, there is a lack of information on the minimum ignition energy (MIE) of organic substances in the literature. On the other hand, MIE is essential information for the proper selection of explosion-proof equipment. Therefore, for application purposes, the MIE prediction model is needed. In this study, based on goodness-of-fit, robustness, predictive capability, and applicability domain (AD), ten quantitative structure-activity relationship (QSAR) models of MIE with different numbers of molecular descriptors were evaluated. A nine-descriptor model was found to have the best performance. The goodness-of-fit performance (R 2 ), robustness (Q 2 Loo ), and predictive capability (Q 2 ) of the proposed model are 0.926, 0.601, and 0.794, respectively. The average absolute error (AAE) of training data and test data is 0.080 mJ and 0.225 mJ, respectively. Compared with the existing QSAR models in the literature, this model has better performance. In addition, the AD of the proposed model is clearly discussed, which is the required element for considering the QSAR model for regulatory application purposes. Highlights: A nine-descriptor QSAR model for predicting the MIE of organic substances is developed. The performance of goodness-in-fit and predictive capability are 0.926 and 0.794. This performances are much better than the performances of existing models. The AD associated with this model was clearly discussed to meet the requirements for regulatory applications.